EP0848318A2

EP0848318A2 - Multi processor system

Info

Publication number: EP0848318A2
Application number: EP97309790A
Authority: EP
Inventors: Ronald Xavier Arroyo Arroyo; Khuong Huu Pham Pham
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 1996-12-10
Filing date: 1997-12-04
Publication date: 1998-06-17
Also published as: EP0848318A3; CN1117327C; CN1184976A; US5802355A

Abstract

A method and apparatus of allowing processors of different speeds to be used in a multi-processor system are disclosed. The method and apparatus comprise a programmable array logic (PAL) or field programmable gate array (FPGA) that detects each of the processors maximum speed and selects a speed common to all of the processors as the operating speed of the processors. The method and apparatus also adjust the system clock to match the speed of the processors.

Description

The present invention relates generally to a multi-processor system, and more particularly, to a multi-processor system employing processors with different maximum speeds of operation.

Many modern computer systems are symmetrical multi-processor systems. In a symmetrical multi-processor system, each processor is used to perform a particular task. For example, each processor may be used to work on a different set of instructions of a process or on a different process altogether, or each processor may be designed and used to handle certain tasks with a high degree of efficiency. As can be surmised, the use of more than one processor in a computer system, often times, enhances the system's performance.

Some computer manufacturers provide symmetrical multi-processor computer systems with replaceable processors. The processors most often reside each on a daughter card that is pluggable onto the motherboard of the computer system via a connector. Thus, when a processor malfunctions or needs to be upgraded, it can be easily replaced by a new processor. The new processor, however, must operate at the same speed as the other processors to comply with the architecture of a symmetrical multi-processor system.

As is well known, the operating speed of processors is ever so increasing. For example, processor speeds that are standard today may well be obsolescent in a year or so.

Hence, replacing a processor with one that operates at the same speed may increasingly be difficult as time goes by. Consequently, there may come a time when replacing a failed processor will necessitate replacing all the other processors in the computer system to ensure the system's operability. Replacing all the processors of a multi-processor system because of a failed processor is a relatively expensive proposition.

Therefore, there is a need in the art for an apparatus and method of using multiple processors of different speeds of operation in a computer system.

The need in the art is addressed by the present invention. The present invention provides an apparatus of allowing processors of different speeds to be used in a multi-processor system. In one embodiment of the invention, the apparatus comprises a programmable array logic (PAL) or field programmable gate array (FPGA) that detects each of the processors' maximum speed and selects a speed common to all of the processors as the operating speed of the processors.

In another embodiment of the invention, the apparatus adjusts the system clock to match the speed of the processors.

How the invention may be carried out will now be described by way of example only and with reference to the accompanying drawings in which:

Fig. 1 is a block diagram of a computer system 100 employed in a preferred embodiment of the invention.

Fig. 2 is a block diagram of the processors and circuitry employed in the present invention.

Fig. 3(a) depicts a first of two logic diagrams used for providing the operating speed of the processors.

Fig. 3(b) depicts a second of the two logic diagrams used for providing the operating speed of the processors.

Fig. 1 is a block diagram of a computer system 100 employed in a preferred embodiment of the invention. Such computer may take the form of a workstation such as the RS/6000 systems marketed by the IBMCorporation, although the invention is not intended to be so limited and is equally applicable to essentially any computer system.

Still referring to Fig. 1, the computer system 100 contains a plurality of central processing units (CPUs) 110 and 120 connected to a system memory 140 through a host bridge 130 on system bus 150. The host bridge is connected to a peripheral component interconnect (PCI) bus 160 having PCI devices 162 - 168 attached thereto. The PCI devices may comprise any of the following: graphics adapters, communication adapters, network adapters, compact disk, floppy disk, hard disk drives etc. The

CPUs

110 and 120 are each connected to the computer system 100 using

connectors

170 and 180, respectively.

Fig.2 is a block diagram of the processors and circuitry employed in the present invention. The circuitry entails a programmable array logic (PAL) 220 connected to both the

processors

110 and 120 via

connectors

170 and 180.

The

connectors

170 and 180 have a plurality of pins. The PAL 220 is also connected to a multiplexer 240. The multiplexer has two inputs: a 66 MHz and a 60

MHz oscillator

250 and 260. The output of the multiplexer 240 is connected to a clock generator 230. The clock generator has a plurality of outputs 270, one of which is connected to the

processors

110 and 120.

According to the preferred embodiment of the invention, four of the plurality of the pins of each of the connectors, presence detect pins 0 to 3 (or PD[0..3]), are used to detect the presence as well as the speed of each of the

processors

110 and 120. The values of the four pins are defined as shown in table 1 below.

PD[0..3]	*CPU WITH MAX. FREQ.*
0000	66 MHz
0001	100 MHz
0010	120 MHz
0011	133 MHz
0100	150 MHz
0101	166 MHz
0110	180 MHz
0111	200 MHz
1111	NO CARD PRESENT

The values of the four pins are provided to the PAL 220 over

signal lines

205 and 215. Normally, the values of the pins from each of the

connectors

170 and 180 would match since both processors would have the same maximum speed. If, however, the values do not match, it is an indication that the

processors

110 and 120 do not have the same maximum speeds. The PAL 220 then selects the slowest of the two speeds as the operating speed of the processors. The PAL 220 also selects the proper frequency of the system clock by providing a signal to the multiplexer 240 over line 235. Once the proper frequency is fed to the clock generator 230, the generator 230 outputs the proper system clock. The system clock is then provided to the

processors

110 and 120 over line 280. Each processor speed chosen by the PAL 220 has associated with it a multiplier that is based on the maximum bus frequency that the system will support. The following two tables illustrate how a particular system would set the multipliers based on the maximum system bus frequency.

		MAXIMUM SYSTEM BUS FREQ. (MHZ)
CPU SPEEDS (MWz)	N_PD [0..3]	50	60 OR 66	66
66	0000	1:1	1:1	1:1
100	0001	2:1	1.5:1	1.5:1
120	0010	2:1	2:1	1.5:2
133	0011	2.5:1	2:1	2:1
150	0100	3:1	2.5:1	2:1
166	0101	3:1	2.5:1	2.5:1
180	0110	3.5:1	3:1	2.5:1
200	0111	4:1	3:1	3:1

N_PD[0..3]	PLL_config[0..3]
0000	0000
0001	1100
0010	0100
0011	0100
0100	0110
0101	0110
0110	1000
0111	1000

The multipliers or ratios are used by the processors to operate at the speed selected by the PAL 220 using the system bus clock. For example, suppose processor 110 has a maximum speed of 133 MHz and processor 120 has a maximum speed of 166 MHz. Once the computer system 100 is turned on or is reset, processor 110 would provide a 0011 signal to the PAL 220 and processor 120 would provide a 0101 signal. The PAL 220 then selects 133 MHz as the operating speed of the two processors. Consequently, the PAL 220 will generate a 0011 signal as the new presence detect bits (i.e., N_PD[0..3]). Using Table 3 as a look-up table, the PAL 220 will send 0100 as the PLL_config[0..3] signal to the processors to configure the phase locked loop (PLL) of each processor. Phase locked loops are used to construct frequency multipliers. U.S. Patent 4,689,581, whose disclosure is incorporated herein, discloses an implementation of a PLL circuit.

As shown in Table 2, this signal instructs the PLL of each processor to multiply the clock signal by 2.5 if the system bus frequency is 50 MHz or by 2 if the bus frequency is 66 MHz. In the present embodiment, since the system bus frequency can either be 60 or 66 MHz, the PAL 220 selects oscillator 250 (i.e., 66 MHz) since the processors will be able to operate closer to the 133 MHz speed using the 66 MHz clock rather than the 60 MHz clock.

The PAL 220 uses the following algorithm to generate the N_PD[0..3] bits:

win0(0) = 1;

win1(0) = 1;

N_PD(0) = PD0(0) & PD1(0);

for (i=1; i<4; i++)

{/* calculate which terms to use to calculate the New PD bits */

win0[i] = win0(i-1) & [ N_PD(i-1) | ( win0(i-1) & !PD0(i-1) ) ];

win1[i] = win1(i-1) & [ N_PD(i-1) | ( win1(i-1) & !PD1(i-1) ) ];

/* calculate the New PD bits */

N_PD(i) = ( !win0(i) | PD0(i) ) & ( !win1(i) | PD1(i) );

};

PD0[0..3] are the presence detect bits of the first processor and PD1[0..3] are those of the second processor. The win0(i) term is associated with the first processor and the win1(i) term is associated with the second processor.

Using the example above, therefore, PD0[0..3] = 0011, PD1[0..3] = 0101 and N_PD(0) = 0 & 0 = 0, the following then occurs:

at i	= 1;
win0(1)	= 1 & [ 0 \| ( 1 & 1 ) ] = 1;
win1(1)	= 1 & [ 0 \| ( 1 & 1 ) ] = 1;
N_PD(1)	= ( 0 \| 0 ) & ( 0 \| 1 ) = 0;

at i	= 2;
win0(2)	= 1 & [ 0 \| ( 1 & 1 ) ] = 1;
win1(2)	= 1 & [ 0 \| ( 1 & 0 ) ] = 0;
N_PD(2)	= ( 0 \| 1 ) & ( 1 \| 0 ) = 1;

at i	= 3;
win0(3)	= 1 & [ 1 \| ( 1 & 0 ) ] = 1;
win1(3)	= 0 & [ 1 \| ( 0 & 1 ) ] = 0;
N_PD(3)	= ( 0 \| 1 ) & ( 1 \| 1 ) = 1;

Thus N_PD[0..3] = 0011.

Fig. 3(a) and Fig. 3 (b) depict logic diagrams used by the PAL 220 to implement the above algorithm when i > 0. As can be seen from the algorithm, two sets of the logic diagram of Fig.3 (a) are used, one for each processor.

In sum, after detecting the operating speed of all the processors of a multi-processor system, the invention selects a speed common to all the processors (ordinarily the speed of the slowest processor) as the operating speed of all the processors. The invention also adjusts the system clock to match the operating speed of the processors. Furthermore, the invention drives the PLL configuration lines of each of the processors in order to provide the correct multiplier. The processors use this multiplier in conjunction with the system clock to operate at the speed selected by the invention.

Consequently, a failed processor of a symmetrical multi-processor system may be replaced by another processor operating at a different speed without replacing all the other processors. In addition, the invention allows the multi-processor system to be gradually updated by replacing the processors one at a time.

The present invention also relates to a method of allowing processors (110, 120) of different speeds to be used in a multi-processor system (100) comprising the steps of:

detecting each processor's maximum speed; and

selecting a speed common to all of said processors (110, 120) as operating speed of said processors.

The method may comprise the further step of adjusting said multi-processor system's clock (230) to match said operating speed and providing to said processors (110, 120) a signal indicative thereof. The method may also comprise providing a multiplier to said processors, said multiplier being used in conjunction with said signal by said processors (110, 120) to function at said operating speed.

The multiplier may be provided by driving proper Phase Locked Loop configuration lines of said processors. The step of adjusting said clock may include the step of selecting one of a plurality of oscillators and the step of providing said selected oscillator to a clock generator.

The detecting and selecting steps may include the step of using a programmable array logic (PAL) or field programmable gate array (FPGA).

Although the present invention has been fully described above with reference to a specific embodiment, other alternative embodiments will be apparent to those of ordinary skill in the art. For example, the PAL 220 may be replaced by a field programmable gate array (FPGA) without departing from the scope of the invention. In addition, the processors need not be on a daughter card, they can be attached to a motherboard using their prongs.

Therefore, the above description should not be taken as limiting the invention's scope which is defined by the appended claims.

Claims

An apparatus for allowing processors (110, 120) of different speeds to be used in a multi-processor system (100) characterised by:

means (170, 180) for detecting each processor's maximum speed; and

means, responsive to said detecting means, for selecting a speed common to all of said processors as operating speed of said processors (110, 120).
A multi-processor system (100) characterised by

means for detecting each processor's maximum speed; and

means, responsive to said detecting means, for selecting a speed common to all of said processors as the operating speed of said processors.
The apparatus or system of Claim 1 or 2 respectively further comprising means for adjusting said multi-processor system's clock to match said operating speed and means for providing to said processors (110,120) a signal indicative thereof.
The apparatus or system of Claim 3 wherein said selecting means comprises providing a multiplier to said processors (110,120) said multiplier being used in conjunction with said signal by said processors to function at said operating speed.
The apparatus or system of Claim 4 wherein providing said multiplier includes driving proper Phase Locked Loop configuration lines of said processors (110,120).
The apparatus or system of Claim 5 wherein said clock adjusting means includes means for selecting one of a plurality of oscillators (250,260) and means for providing said selected oscillator to a clock generator (230).
The apparatus or system of Claim 6 wherein said detecting and selecting means include using a programmable array logic (PAL) (220) or field programmable gate array (FPGA).
A method of allowing processors (110,120) of different speeds to be used in a multi-processor system (100) comprising the steps of:

detecting each processor's maximum speed; and

selecting a speed common to all of said processors (110,120) as operating speed of said processors.
The method of Claim 8 further comprising the step of adjusting said multi-processor system's clock (230) to match said operating speed and providing to said processors (110,120) a signal indicative thereof.
The method of Claim 9 wherein said step of selecting comprises providing a multiplier to said processors, said multiplier being used in conjunction with said signal said processors (110,120) to function at said operating speed.